Correlation between the colors of body areas

Analysis of the colors of body areas, no recoding.

Mutual information between colors in body areas

This cannot tell apart correlation from anticorrelation.

library(infotheo)
library(ggplot2)
library(reshape2)

bodyAreas <- d[,3:26]
names(bodyAreas)<-as.character(c(1:24))
# Remove NAs
bodyAreas<-na.omit(bodyAreas)

qplot(x=Var1, y=Var2, data=melt(mutinformation(bodyAreas)), fill=value, geom="tile") + labs(x="Body areas", y="Body areas")

Basic Pearson correlation between colors in body areas

This is not correct as we are dealing with a categorical variable.

sp<-qplot(x=Var1, y=Var2, data=melt(cor(bodyAreas)), fill=value, geom="tile") + labs(x="Body areas", y="Body areas")
sp+scale_fill_gradient2(midpoint=0, low="blue", high="red", mid="white", space="Lab")

Recoding in 0/1 depending on black/non black.

Mutual information

library(infotheo)
bodyAreasBW <- bodyAreas
bodyAreasBW[bodyAreasBW>0]<-1


qplot(x=Var1, y=Var2, data=melt(mutinformation(bodyAreasBW)), fill=value, geom="tile") + labs(x="Body areas", y="Body areas")

Pearson correlation

Here it is somewhat correct as we have recoded in 0/1.

correlations <- cor(bodyAreasBW)

sp2<- qplot(x=Var1, y=Var2, data=melt(correlations), fill=value, geom="tile") + labs(x="Body areas", y="Body areas")

sp2+scale_fill_gradient2(midpoint=0, low="blue", high="red", mid="white", space="Lab")

Summary of the correlations.

summary(c(correlations))

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
## -0.1229  0.0185  0.1151  0.2465  0.4144  1.0000